Exploring Functional Temporal Data

P. Montero Manso

The statistical analysis of temporal data from an infinite-dimensional perspective is shown to be beneficial yet challenging, creating a demand for tools to explore such wealth of complex data and methods. We propose a tool that works on two levels: First, a large collection of data sets is processed using methods such as two sample testing,
classification or regression, creating a reference database of data, methods and results. Second, techniques from clustering and visualization, such as distances and feature extraction for temporal data are provided to easily query and explore the database. Using this tool, a practitioner facing new, unstudied data can link it to well analyzed problems in the database via different notions of similarity, thus finding methods that will work well for the new data. In a similar fashion, statisticians may recover suitable data to apply their methodologies and compare them with existing approaches. The tool is implemented as an R package, with examples.

Palabras clave: functional data time series database visualization data mining

M. D. Ruiz Medina, F. D. Miranda Huaynalaya, R. M. Espejo Montes

G. Mestre Marcos, J. Portela González, A. Muñoz San Roque, E. Alonso Pérez

R. E. Lillo Rodríguez, A. Franco Pereira

A. Elías Fernández, R. Jiménez Recaredo